Mid-Holocene Social Interaction in Melanesia: New Evidence from Hammer-Dressed Obsidian Stemmed Tools

Home , Biak, Bougainville Island, Melanesia, New Guinea, Pam Islands, Trobriand Islands

ROBIN TORRENCE, PAMELA SWADLING, NINA KONONENKO, WALLACE AMBROSE, PIP RATH, AND MICHAEL D. GLASCOCK

introduction Proposals that large-scale interaction and ceremonial exchange in the Pacific region began during the time of Lapita pottery (c. 3300–2000 b.p.) (e.g., Friedman 1981; Hayden 1983; Kirch 1997; Spriggs 1997) are seriously challenged by the extensive areal distribution of a class of retouched obsidian artifacts dated to the early and middle Holocene (c. 10,000–3300 b.p.) and known as ‘‘stemmed tools’’ (Araho et al. 2002). Find spots of obsidian stemmed tools stretch from mainland New Guinea to Bougainville Island and include the Trobriand Islands, various islands in Manus province, New Britain and New Ireland (Araho et al. 2002; Golson 2005; Specht 2005; Swadling and Hide 2005) (Fig. 1). Although other forms of tanged and waisted stone tool are known in Melanesia (e.g., Bulmer 2005; Fredericksen 1994, 2000; Golson 1972, 2001), the two types defined by Araho et al. (2002) as ‘‘stemmed tools’’ comprise distinctive classes because they usually have deep notches that delineate very well-defined and pronounced tangs. Type 1 stemmed tools are made from prismatic blades and have large and clearly demarcated, oval-shaped tangs. In contrast, the Type 2 group is more vari- able.Itisdefinedprimarilybytheuseof Kombewa flakes (i.e., those removed fromthebulbarfaceofalargeflake)forthe blank form, as described in detail in

Robin Torrence is Principal Research Scientist in Anthropology, Australian Museum, Sydney NSW, [email protected]; Pamela Swadling is a Visiting Research Fellow, Archaeol- ogy and Natural History, Research School of Pacific Studies, Australian National University, Can- berra ACT, [email protected]; Nina Kononenko is an ARC post-doctoral fellow in the School of Philosophical and Historical Inquiry, University of Sydney, kononenko.nina@hotmail. com; Wallace Ambrose is a Visiting Research Fellow in Archaeology and Natural History, Research School of Pacific Studies, Australian National University, [email protected]; Pip Rath is a post-graduate student in the School of Philosophical and Historical Inquiry, University of Sydney, Australia, [email protected]; Michael D. Glascock is Senior Research Scientist, Research Reac- tor Center, University of Missouri, Columbia, MO, USA, [email protected]. Asian Perspectives,Vol.48,No.1( 2009 by the University of Hawai‘i Press. Fig. 1. Map showing the location of obsidian stemmed tools. With the exception of Garua Island, Willaumez Peninsula, and Mopir where numerous stemmed tools have been found, the symbols represent a single find spot. torrence et al. . mid-holocene social interaction in melanesia 121

Araho et al. (2002). Since Kombewa technology is extremely rare in stone tool assemblages from Melanesia, this trait alone easily separates Type 2 stemmed tools from other retouched tools. The diagnostic characteristics of Type 1 and 2 stemmed tools with lengths generally greater than 750 mm clearly distinguish these tool types from other waisted or tanged chipped stone tools in this region. Examples of obsidian stemmed tools, mostly found in surface contexts, were reported decades ago (cf. Bulmer 2005; Golson 1972, 2001) but, lacking firm chronological context, their significance was not appreciated until relatively large numbers were recovered from stratified and well-dated deposits in recent excavations in New Britain, Papua New Guinea (e.g., Specht et al. 1988; Torrence et al. 1990). Their chronological assignment to the early and middle Holocene is based on their consistent recovery from radiocarbon-dated soils or from deposits within a well dated tephrastratigraphy and their absence in contexts dating to other peri- ods (Araho et al. 2002:62). In New Britain large stemmed tools are found in soils buried underneath either the radiocarbon-dated W-K1 volcanic tephra (c. 6160– 5740 b.p.) or the W-K2 tephra (3480–3160 b.p.) (Petrie and Torrence 2008). In contrast, Lapita pottery is always stratified above and therefore younger than the W-K2 tephra. Araho et al. (2002), Torrence (2003, 2004a), and Specht (2005) suggested that these large stemmed tools functioned as ceremonial valuables ‘‘to create and maintain social links between and within groups and to mark status di¤erences’’ (Araho et al. 2002:77). This hypothesis is based on the use of manufacturing techniques that require considerable e¤ort, skill, and care, and a raw material easily associated with specific, localized places, as well as the distinctive shapes, shininess, symmetry, and fragility of the tools. More recently, Specht (2005), Swadling and Hide (2005:307), and Pavlides (pers. comm.) have reported additional examples. Together with Golson (2005), all these scholars have also proposed that the distributional data of obsidian stemmed tools, shell adzes, and mortars and pestles provide considerable tangible evidence for wide-scale social interaction during the mid-Holocene, long before the existence of Lapita pottery and colonization into Remote Oceania (cf. Torrence and Swadling 2008). Little is known about the social relationships that have been proposed as re- sponsible for the impressive spatial distribution of the stemmed tools. Did these artifacts circulate within a system of ceremonial exchange on the Willaumez Pen- insula, where most examples have been recorded (Fig. 1)? Were the far-flung casesmerelyrareleakagesfromthatsystem or was the entire region where they occur linked through social interaction, and if so, in what ways? We address these questions through a series of analyses stimulated by a newly reported obsidian stemmed tool from Biak Island in West Papua (Friede 2005:I, 41; II, 86) (Fig. 2). Research focused on three unusual characteristics of the Biak artifact has signifi- cantly altered our understanding of stemmed tools. Firstly, geochemical sourcing techniques identified the raw material as derived from obsidian outcrops in Manus rather than West New Britain where stemmed tools are common finds. Secondly, the highly unusual technique of hammer-dressing was used in its manufacture. Thirdly, the Biak provenience greatly extends the spatial distribution of stemmed tools. We consider the significance of these findings for social interaction during the mid-Holocene period and for Pacific prehistory in general. 122 asian perspectives . 48(1) . spring 2009

Fig. 2. Obsidian stemmed tool found on Biak: (left) hammer-dressing is visible both on the stem and on the middle of the blade; (right) hammer-dressing has obscured most of the original surface of the ﬂake, but a portion is visible on the top left. Scale bar is 3 cm. (Photo by R. Torrence)

biak tool The Biak tool clearly fits the criteria for a Type 2 obsidian stemmed artifact because it was made from a large Kombewa flake and bears a prominent tang created by deep notches (Fig. 2). At 132 mm in length, it is within the range of the large Type 2 tools proposed as valuables (cf. Araho et al. 2002; Specht 2005:381; Torrence 2003, 2004b). Although the tool does not come from a secure archaeological context, we believe that the distinctive similarity with the well-dated obsidian stemmed tools from New Britain is close enough to propose that the Biak tool belongs to the same typological group and therefore also dates to the mid-Holocene. The stem on the tool was first shaped by bifacial flaking on what was probably the proximal end of a large Kombewa flake. Irregularities created by the flaking were then partially removed by hammer-dressing. The direct, hard hammer blows created a uniformly even, but rough-textured, surface with numerous, small impact points (Fig. 3). Hammer-dressing has obscured flaking on the stem, except for several flat scars at the base. The hammer-dressing was also applied on both sides of the tool to surfaces that were originally smooth and shiny. On one flake surface the hammer-dressing is continuous with that on the stem (Fig. 2, right), torrence et al. . mid-holocene social interaction in melanesia 123

Fig. 3. Close-up of hammer-dressing on the stem of the Biak tool. Scale bar is 2 cm. (Photo by R. Torrence) whereas on the other there is an isolated, circular patch of hammering on the blade of the tool (Fig. 2, left). As indicated by flake scars with di¤erent degrees of weathering, the tool has been extensively reworked. Areas of retouch around the entire working edge include two large, invasive, fresh scars struck from opposing sides, which stretch across much of the surface of one face (Fig. 2; right; cf. Specht 2005: Figs. 28.4, 28.5). John Friede (2005: II, 86), a private collector who kindly loaned the artifact to Swadling for analysis, reports that the new stemmed tool was ‘‘found in a cave on Biak Island with Japanese war relics.’’ We have no information about whether the tool was excavated or was found on the surface. It seems highly unlikely that the tool had been transported to the island by a Japanese soldier. Biak had been of 124 asian perspectives . 48(1) . spring 2009 little strategic importance to the Japanese until faced with the westward advance of the Allied forces along the coast of New Guinea. The island has no natural harbors and fresh water is not readily available. In an e¤ort to stop the advancing Allied forces, the Japanese decided to make Biak a key air base, as it was within fighter range of many existing air bases in Dutch New Guinea. Late in 1943, they began constructing airstrips in the southeastern part of the island on the coastal plain. The troops sent to protect and hold Biak did not come from the Pacific, but were sent from China (Smith 1953:280, 281, 299). Sketches by an American soldier show a range of defensive walls constructed in the Biak caves (Riegelman 1955). It seems likely that in building these, any archaeological deposits would have been highly disturbed. Given the historical data to hand, the most parsimonious assumption is that the Biak stemmed tool was deposited in the cave during prehistory and was uncovered during the construction of defenses by the Japanese or, perhaps, through more recent activity. Regardless of the lack of stratigraphic integrity for the Biak tool, its strong resemblance to Type 2 stemmed tools made on Kombewa flakes provides strong support for accepting a mid- Holocene date for its manufacture.

characterization studies PIXE-PIGME analyses have demonstrated that obsidian from all the major obsidian sources in New Britain (Gulu, Kutau-Bao, Baki, Mopir) was used in the production of stemmed tools (Araho et al. 2002; Rath and Torrence 2003) and that three stemmed tools in the Sepik-Ramu region of the New Guinea mainland were also made from New Britain obsidian (Specht 2005:384; Swadling and Hide 2005:307; Figs. 1, 4). Since the Biak stemmed tool is very similar in form

Fig. 4. Location of New Britain, Lou Island, and Pam Island obsidian sources. torrence et al. . mid-holocene social interaction in melanesia 125

Table 1. Hammer-dressed artifacts: Types and sizes

stem maximum total stem length/total stem stemmed tool type length length length thickness

Biak tool 2, Kombewa, reworked 132 75 0.57 33 FAP M200 2, Kombewa 107 69 0.64 28 FAP M267 2, Kombewa 93 44 0.46 16 FAP M408 2, Kombewa 77 45 0.58 35 FAP M409 2, Kombewa 89 55 0.62 37 FAP M702 2, Kombewa, stem absent FAP M224 2, broken stem only 28 FAP M228 Noncortical flake FAO 1000/1010 5.1 Cortical flake Lihir Island 2?, secondary flake, (New Ireland) stem absent

Notes: Measurements in mm. Sites FAP and FAO are located on Garua Island. For locations see Figure 1.

Table 2. Hammer-dressed stemmed tools: Shapes and profiles

longitudinal stemmed tool notching blade shape working edge section

Biak tool Moderate Rectangular Straight Rectangular FAP M200 Weak Trapezoidal Slightly convex Lenticular FAP M267 Moderate Rectangular Slightly sinuous Rectangular FAP M408 Weak Triangular Straight Lenticular FAP M409 Weak Triangular Straight Lenticular FAP M702 ? Semi-circle Convex Lenticular Lihir Island ? Oval Convex, slightly ? rectangular (New Ireland) pointed

and manufacturing technique to these artifacts, one could reasonably infer that it was also derived from New Britain. To test this assumption, we employed a combination of PIXE-PIGME, LA/ICPMS, and INAA techniques to characterize the Biak tool and samples from six stemmed tools and two ﬂakes from Garua Island that also have traces of hammer-dressing (Tables 1 and 2; Figs. 5–7). Ambrose obtained su‰cient material for the geochemical study and then conserved the Biak artifact so the sampling region was invisible. The ﬁrst analysis used LA/ICPMS at the Australian National University to characterize the Biak artifact and two stemmed tools from site FAP on Garua Island. The results are presented in Tables 3 and 4 where they are compared to source samples. A correspondence analysis of the Papua New Guinea obsidian sources shows that the Biak tool clusters with source material from Lou and Pam Islands in Manus province (Fig. 8). In contrast, the stemmed tools from site FAP on Garua Island match the composition of sources from the Baki source, also on Garua Island. The East Fergusson cluster comprises samples from Dobu Island, 126 asian perspectives . 48(1) . spring 2009

Fig. 5. Obsidian stemmed tools from Garua Island: upper, FAP M409; lower, FAP M408. Scale bar is 3 cm. (Photos by R. Torrence)

Sanaroa Island, and Lamonai; the Southwest Manus obsidians are surface finds clearly distinct from any other group; the West Fergusson group includes obsidians from Igwageta and Iaupolo but Fagalulu is chemically di¤erentiated; and, finally, the Talasea (Kutau-Bao) and Voganakai (Gulu) groups both comprise several exposures that can be di¤erentiated, but are clustered in this analysis. Moving to a finer level of discrimination, the correspondence plot of ICPMS results in Figure 9 demonstrates that the Biak artifact is most closely aligned with the obsidian sources found in deposits within the Umleang-Umrei region on Lou Island (Ambrose et al. 1981). All the samples are from Lou Island, except those Fig. 6. Obsidian stemmed tools from Garua Island: upper, FAP M200; lower FAP M702. Scale bar is 3 cm. (Drawing by N. Kononenko)

Fig. 7. Obsidian stemmed tool from Garua Island, FAP M267, showing location of points where use-wear is illustrated in Figures 10 and 11: A. ?ventral side of the Kombewa ﬂake; B. ?dorsal side of the Kombewa ﬂake. Scale bar is 3 cm. (Drawing by N. Kononenko) 128 asian perspectives . 48(1) . spring 2009

Fig. 8. Correspondence analysis of obsidian sources from Papua New Guinea using measurements of V, Co, Y, Nb, Mo, Cs, Er, and Ta obtained from LA/ICPMS. The plot compares results from the ﬁrst two factors in the analysis.

Table 3. Summary of results of characterization studies of hammer-dressed artifacts stemmed tool la/icpms pixe-pigme inaa

Biak tool Umleang, Lou Island Umleang, Lou Island FAP M200 Baki FAP M408 Baki Baki FAP M409 Baki Baki FAP M228 Kutau FAO 1000/1010 5.1 Baki

Note: For location of obsidian sources see Figure 4. from the Pam Lin and Pam Mandian Islands nearby. The Wekwok results represent a standard taken from a single block of obsidian. They provide an indication of the degree of variation within the current LA/ICPMS system. To substantiate and expand on these results, the same samples used in the LA/ ICPMS study, together with samples taken from two additional hammer-dressed Table 4. Element concentrations for Baki (Garala) and Lou Island/Umleang obsidian sources (Ã) and for stemmed tools using LA/ICPMS analysis sample V Co Y Zr Nb Mo Cs Er Ta

Biak tool 4.19 1.73 33.10 345 47.6 3.85 2.00 3.67 3.27 ÃLou Mine 4 ANU1131 4.44 1.79 33.00 344 47.4 3.87 1.93 3.72 3.25 ÃLou Mine 22 ANU5232 4.43 1.86 33.20 346 47.7 3.89 1.96 5.69 3.26 FAP M408 WNB 3529 2.93 0.48 22.11 120 2.21 3.03 1.42 2.52 0.14 FAP M409 WNB 3528 2.72 0.45 20.68 112 2.06 2.87 1.34 2.33 0.13 ÃBaki ANU2365 2.97 0.49 23.12 124 2.16 3.02 1.45 2.57 0.13 ÃBaki ANU2367 2.95 0.48 23.51 125 2.15 2.97 1.45 2.60 0.14

Note: ANU and WNB designate samples taken from artifacts.

Fig. 9. Correspondence analysis of obsidian sources from the Lou-Pam Islands cluster in Figure 7 based on measurements of V, Co, Y, Nb, Mo, Cs, Er, and Ta obtained from LA/ICPMS. The plot compares results from the ﬁrst two factors in the analysis. 130 asian perspectives . 48(1) . spring 2009

Table 5. Element concentrations of stemmed tools using PIXE-PIGME analysis. WNB designates samples taken from artifacts

F Na Al Si K Ca Ti Mn Fe Rb Sr Y Zr Nb sample ppm % % ppm % % ppm ppm % ppm ppm ppm ppm ppm

WNB 3527 1264 3.33 7.07 34 3.61 1.03 2164 553 2.16 159 78 47 427 45 Biak tool WNB 3528 647 2.63 6.34 38 3.65 0.83 1511 505 1.25 65 164 33 179 4 FAP M409 WNB 3529 615 2.63 6.32 37 3.66 0.83 1529 522 1.27 72 164 33 181 0 FAP M408 WNB 1819 625 2.64 6.19 35 3.45 0.77 1418 492 1.21 60 149 29 187 2 FAP M200 WNB 2260 555 2.65 6.43 40 3.60 0.97 1642 533 1.08 58 212 17 152 0 FAP M228

Note: WNB designates samples taken from artifacts.

Table 6. Element concentrations for obsidian source groups using PIXE-PIGME analyses

F Na Al Si K Ca Ti Mn Fe Rb Sr Y Zr Nb source no. ppm % % ppm % % ppm ppm % ppm ppm ppm ppm ppm

Umrei/Umleang 32 1329 3.39 6.96 32 3.54 1.02 2183 626 2.14 159 77 37 448 50 (Lou Island) s.d. 48 0.15 0.24 2 0.2 0.05 121 30 0.1 8 5 5 18 6 Baki 55 637 2.59 6.29 36 3.71 0.82 1500 509 0.82 60 146 29 179 2 s.d. 16 0.12 0.1 2 0.29 0.04 87 21 0.06 3 4 3 27 2 Baki (Garala C) 10 620 2.56 6.15 32 3.36 0.74 1390 455 1.07 55 129 26 153 1 s.d. 29 0.18 0.08 3 0.39 0.05 100 31 0.06 4 0 3 10 2 Kutau-Bao 47 559 2.67 6.34 37 3.46 .95 1600 523 1.05 55 206 20 151 1 s.d. 12 .08 .09 2 .20 .05 95 22 0.05 4 9 2 7 2

Note: All data based on Bird et al. (1997) except for Umrei/Umleang (Lou Island), which is derived from Bird (1996). artifacts from Garua Island were submitted to PIXE-PIGME analysis at ANSTO using the basic protocol described in Summerhayes et al. (1998). The PIXE- PIGME determinations conﬁrm the LA/ICPMS results (Tables 3, 5, 6). Based on extensive studies of obsidian sources (Bird 1996; Bird et al. 1997), the Biak tool can be assigned to the Umleang/Umrei source on Lou Island, whereas all of the hammer-dressed stemmed tools from Garua Island are associated with the Baki chemical group, some of which outcrops in the same streambed where the tools were found. FAP M200 was assigned to the GaralaC subgroup of the Baki chemical group, which outcrops both on Garua Island near the site of FAP and on nearby Garala Island. Interestingly, the non-cortical ﬂake FAP M228 groups with the Kutau-Bao obsidian source located on the New Britain mainland, c. 1 km away (Fig. 4). Finally, instrumental neutron activation analysis (INAA) at the torrence et al. . mid-holocene social interaction in melanesia 131

Table 7. INAA analysis of nine source samples and one cortical flake with hammer-dressing from the FAO site (NAA 14) sample Dy (ppm) Mn (ppm)

NAA 14 4.39 475 Baki source mean 4.59 482 Baki source s.d. 0.34 5.77

University of Missouri Research Reactor (MURR), using methods described in Glascock et al. (1998), found that a further cortical ﬂake with hammer-dressing matched the composition of the Baki source (Tables 3, 7). The characterization studies provide the ﬁrst evidence that stemmed tools were manufactured using obsidian from Manus as well as from New Britain. Before considering the profound social implications of Manus as a second source for stemmed tools (cf. Specht 2005:385), we turn to the second unique attribute of the Biak tool.

an unusual technique The use of hammer-dressing on the stem and blade of the Biak tool is not recorded for flaked obsidian artifacts in any other areas of the world. Grinding, another technique that does not involve flaking, is found on the platforms of obsidian blade cores from Mesoameria (e.g., Sheets 1978), but hammer-dressing has not been reported previously on flaked obsidian tools, although it is common on artifacts made from less isotropic stone. In contrast, this highly unusual technique for flaked artifacts is found on rare stemmed tools from Garua Island (Fig. 1), but is absent on the neighboring New Britain mainland, even from large assemblages at sites FRL and FCH (Araho 1996; Araho et al. 2002; Fullagar 1992; Specht 1973, 1974; Specht et al. 1988), or at the Mopir obsidian outcrops (Fullagar et al. 1991). Outside New Britain, hammer-dressing occurs on a broken retouched flake from Lihir, New Ireland (Casey 1939:149–150, fig 7b; Specht 2005:384), which may part of a Type 2 stemmed tool made on a ‘secondary series’ Kombewa flake, as defined by Araho et al. (2002:66, figs. 7, 12; Fig. 1). Among the tens of thousands of artifacts examined from Garua Island, only eight bear hammer-dressing (Tables 1 and 2): four complete and one broken Type 2 stemmed tools; one bifacially retouched stem that might derive from a broken Type 2 form; and two flakes (Figs. 5–7). All the artifacts are derived from the FAP obsidian quarry and workshop site, which is located adjacent to the Baki obsidian source (e.g., Torrence et al. 1990:461) or were excavated from site FAO situated on a hilltop directly above the obsidian outcrops. Since the Garua hammer-dressed tools are in association with debris from tool manufacture, they may be rejects rather than finished examples. This might explain why the Garua artifacts are smaller than the Biak tool, although the proportions of the stem and blade are similar (Table 1). On the other hand, the Garua examples could be tools that had been used, but were thrown away once replacements had been made at the quarry, or tools used in activities within this general location. 132 asian perspectives . 48(1) . spring 2009

The two waste flakes demonstrate that hammer-dressed tools were manufactured on Garua Island (Table 3). Given the type of force applied during hammer- dressing, breakages obviously occurred during manufacture. For example, only a small pecked area is present on the FAP M224 broken stem, which implies that during manufacture the tool snapped cleanly at the weakest point: the junction of the stem and the rest of the flake. Like the Biak tool, the Garua hammer-dressed stemmed tools have a robust, sturdy feel to them. Firstly, the stems are quite thick compared to others reported (Table 1; Torrence 2004a :167). Secondly, the notches forming the stem are relatively shallow and any corners created by them are quite thick. Thirdly, although several have a slightly convex working edge viewed in plan, in most cases the edges are relatively straight and di¤er from the highly rounded, crescent shape of other more elaborate types (Araho et al. 2002; Torrence 2003). FAP M267 (Fig. 7) varies from the others because of the presence of retouch along one edge. This was applied to the core before the Kombewa flake that forms FAP M267 was struck (cf. Araho et al. 2002: fig. 8). The Garua hammer-dressed tools are dated by a combination of tephrastratigraphy and AMS dating. FAP M267 (Fig. 7) was found in a stream section imme- diately under a layer of redeposited W-K1 tephra. The unit is dated by an AMS determination of 5204 e 85 (NZA 1570; 5740–6210 b.p.) on charcoal from the same stratigraphic level in a test pit c. 100 m upstream (Torrence et al. 1990:461). Secondly, a cortical flake with hammer-dressing was recovered from a level dated by AMS to 3532 e 66 (NZA 2901; 3640–3990 b.p.) at site FAO. These dates place hammer-dressed stemmed tools on Garua Island firmly within the mid- Holocene and prior to the local appearance of Lapita pottery (Torrence and Stevenson 2000). The presence of hammer-dressing on the Biak and Garua artifacts is intriguing. Obsidian is notable for its shininess, a characteristic often associated with aesthetic and symbolic meanings typical of valuables (cf. Torrence 2005). It therefore seems strange that the tool’s reflective surface was deliberately destroyed. To understand why hammer-dressing was applied, we consider its potential role for tool-use and the wider social implications of this unusual technique.

tool-use Since they are so rare, one might predict that the hammer-dressed tools were special in some way. The plans of stemmed tools with hammer-dressing can be grouped into two broad shapes. Only one tool from Garua (FAP M267) has the same rectangular-shaped blade as the Biak tool (in its current reworked form), whereas the majority has a trapezoidal shape with expanding sides that maximize the length of the working edge. The Lihir tool may represent a third oval, blade shape. The rectangular blades have deeper notches than the expanding blade group. The longitudinal sections of the two groups also di¤er. The Biak and FAP M267 tools have a relatively straight, flat longitudinal profile with a relatively consistent thickness along the whole length of the tool (Figs. 2, 7). Those in the second group have a very thick stem in comparison to the working edge so the profile either tapers smoothly toward the edge (e.g., FAP M702, Fig. 6) or there is a prominent angle along one of the faces (e.g., FAP M200, Fig. 6). In both cases torrence et al. . mid-holocene social interaction in melanesia 133 the result is a relatively thick and long working edge that is controlled from a much narrower stem that could fit nicely within a handle or straight into the hand. The hammer-dressed tools were designed on di¤erent principles than the more elaborate and fragile Type 2 stemmed tools previously proposed as valuables (e.g., Araho et al. 2002; Torrence 2004a). Firstly, the latter usually have a crescent- shaped blade with the rounded edge profile forming an arc. Secondly, there are deep notches leading to fragile protrusions or tips (Araho et al. 2002: figs. 2, 3, 12; Torrence 2003: fig. 18.3, 2004a:figs.3,4).Thirdly,thestemoftheelaborate type is not located at the proximal end of the Kombewa flake. Instead, the bulb of percussion has been removed by the notch which forms the stem. This means that the distal end of the stem is the original sharp, thin edge of the flake (Tor- rence 2004a: Figs. 5, 6). Such thin, fragile stemmed tools would have broken under minimal stress. In comparison, stemmed tools proposed by Torrence (2003, 2004a) to have had utilitarian or mixed uses were made with the thick, bulbar part of the flake forming the distal end of the stem as in the case of the hammer- dressed tools.

Use-Wear Analysis Although the hammer-dressed tools appear to be quite robust and functional, it is still possible that they were used in activities that had special significance. Unfor- tunately, the original edges of almost all have been removed, as in the case of the Biak tool, or extensively damaged by movement within the streambed where they were found. Kononenko found that only FAP M267, which was found in situ under a thick layer of redeposited W-K1 tephra within a steep bank of Malaiol Stream (Torrence et al. 1990:461), was well enough preserved for use- wear analysis. Fortunately, of all the hammer-dressed tools from Garua, FAP M267 most closely resembles the Biak tool in terms of the shape of the blade and the distribution of hammer-dressing. After cleaning, Kononenko examined the artifact under both low- and high-power magnification following the basic meth- odology outlined by Fullagar (2005). A combination of edge morphology, striations, polish, and residues was used to interpret past tool use. The working edge of FAP M267 is slightly rounded in plan view and is still relatively sharp. Under low-power magnification feathered and isolated microscars were observed on both faces along the distal edge (i.e., opposite the stem). More intensive damage, in the form of continuous feathered microscars, is located on the right-hand side of the face in Figure 7B. Comparison of the use-wear microscars with fresh scarring and unused surfaces together with the presence of striations within the microscars indicate that they were formed as a result of tool use. Two forms of striations, (1) narrow and (2) deep and intermittent (Hurcombe 1992:37), were observed mainly on the right-hand side of the face shown in Fig- ure 7B. They run parallel or in a slightly diagonal direction to the edge and some- times cross each other. The striations called ‘‘sleeks’’ located along the edge are primarily observed in association with isolated microscars (Fig. 7, points 1, 2, 3, 5; Fig. 10 A, B, D, F). This pattern of striations was also observed on experimental tools used for processing soft, elastic materials such as skin, meat or fish (Fig. Fig. 10. Use-wear on hammer-dressed stemmed tool FAP M267. For location of points see Figure 7. A, point 1: slightly rounded edge; light polish; discontinuous bending and feather scars; diagonal, deep, isolated striations of sleek and rarely intermittent types (100Â). B, point 2: feather and bending scars; parallel, isolated, long, and deep striations with a few diagonal striations (100Â). C, point 2: intensive edge rounding and developed polish with some diagonal, isolated striations (500Â). D, point 3: feather and bending scars; isolated, deep, parallel, and diagonal striations of intermittent and sleek types. Note that striations spread into scars (100Â). E, point 4: intensive edge rounding; patches of dense, deep, isolated sleek and intermittent striations and a few flaked striations on the edge; developed polish (100Â). F, point 5: intensive edge rounding; discontinuous feather scars; light to developed polish; isolated shallow and deep, parallel striations (100Â). (Photos by N. Kono- nenko) torrence et al. . mid-holocene social interaction in melanesia 135

11E, F). In addition, numerous and more obvious intermittent striations are associated with an area of intensive scarring at point 4 in Figure 7. These wear forms suggest that this part of the edge had more intensive contact with a harder material such as bone. In terms of polishes, there are a few spots with thin lines or patches that have smoothed bright polish that extends from the higher peaks into low areas of the obsidian surface (Fig. 10C). The distributionofthepolish,itsassociationwith edge rounding, and the pattern of the striations provide additional support for the hypothesis that the tool was used for cutting a soft, elastic material, but also occasionally made contact with a harder substance. It is important to note that the degree of use-wear on FAP M267 is relatively intensive. Based on experiments, it was used in excess of 30 minutes. The size and shape of the working edge of FAP M267 indicate its appropriateness for use on large-size animals. The use-wear traces are entirely consistent with cutting the carcass or working the ﬂesh and bones of large-bodied animals such as crocodiles, turtles, tuna, dolphins, or even humans. At this stage the use-wear analysis neither supports nor challenges the possible ceremonial or ritual use of FAP M267.

Hafting The use-wear results suggest that the extra care and attention directed to the handle in the form of hammer-dressing may have been necessary for tool function, because a precision grip may have been required to e¤ectively complete the task.Certainly,ifmuchforcewasapplied,e.g.,inbutcheryorwarfare,asecure hold would have been highly beneficial. Araho et al. (2002:70–71) noted that hammer-dressing converted smooth obsidian into a rough and irregular surface that would better support a strong grip when held directly in the hand or aid in the attachment of a wrapped or wooden handle. The use-wear study found microscopic evidence for hafting on the stem of FAP M267. Along the left edge of the stem in Figure 7B, there are a few areas with scars that are associated with a net of intermittent and sleek striations (Fig. 10A). These are further accompanied by light polish and some plant tissue (Fig. 7, point 20). A similar constellation of abrasion, polish, striations and plant tissue was observed at some elevated areas of the stem (e.g., Fig. 7, point 7; Fig. 11B). These use-wear traces indicate that the haft had come in contact with wood or perhaps from string made from a woody plant, such as rattan. A similar wear pattern was observed on an experimental tool that was placed inside a wooden handle (Fig. 11C, D). The role of hammer-dressing in improving the grip seems particularly apt for tools FAP M267, FAP M200, and FAP M408 since much of the percussion damage occurs on flat, unflaked surfaces (Figs. 5–7). FAP M702 is hammer-dressed on both the flaked surface that forms the notches and on the smooth surface of the flake between them (Fig. 6). It is plausible that the e¤ective application of force on the long working edge on these tools demanded a sturdy handle and this was assisted by the simultaneous smoothing of the bifacially flaked stem and the roughening of the glassy flake surface. The Biak tool may incorporate elements of both these functions because the bifacially flaked stem is extensively covered with hammer-dressing and where this Fig. 11. Hafting wear on hammer-dressed stemmed tool FAP M267 (A, B); hafting wear on experimental tools 178 and 247 (C, D); and use-wear on working edges of experimental tools 263 and 308 (E, F): A. FAP 267 point 6: rounded edge; rare isolated and random oriented deep striations; light polish; few scars covered by plant tissue (100Â); B. FAP M267 point 7: spot of light polish with ran- domly oriented, isolated striations, and plant tissue (100Â). C. experimental tool 178, cutting banana leaves and stem for 30 minutes: hafting wear, rounded edge with developed polish; few isolated striations and plant tissue (200Â). D. experimental tool 247, sawing frond, Nipa palm for 15 minutes, hafting wear: patches of dense, deep striations oriented perpendicularly to the edge and white- colored residues (100Â). E. experimental tool 263, gutting/cutting fish for 25 minutes: light edge rounding; feather and bending scars; light to developed polish; diagonal and parallel, isolated striations (100Â). F. experimental tool 308, gutting/cutting fish for 30 minutes: intensive edge rounding and developed polish; few diagonal, deep striations (500Â). (Photos by N. Kononenko) torrence et al. . mid-holocene social interaction in melanesia 137 is absent, the flake scars are very flat. The impact cones from percussion, however, also extend quite far onto the smooth, shiny, flat flake surface above the notching on both faces of the tool (Fig. 2), suggesting that the roughening of the smooth surfaces may have been applied to secure wrapping or a haft for a handle. The hy- pothetical use of wrapping is very similar to the way long obsidian blades were recently attached to spears in Manus (Moseley 1877: pl. xx; Torrence 1993:472), although there is no evidence of hammer-dressing on these tools. The isolated patch of hammer-dressing on the blade may have had an entirely di¤erent function and history (Fig. 2, left). One possible explanation is that the tool was used as an anvil, perhaps for bipolar working of stone tools, as has been hypothesized for similar patterns on ground edge axes in Australia (McCarthy et al. 1946:44, Figs. 257, 279), although the Australian examples often have a definite pitted area which is not present on the Biak tool.

non-utilitarian roles

Although preparing the obsidian surface so that it can be comfortably and safely held or securely hafted makes good functional sense, hammer-dressing is so rare in Melanesian obsidian assemblages, even among stemmed tools, that we question whether its application on the mid-Holocene obsidian tools described here was purely or even primarily utilitarian. An alternative hypothesis is that the rough- ened surface was aesthetic or had ideological significance. Since obsidian is natu- rally shiny, a dull surface could have carried symbolic meaning. Another possibility is that creating a rough exterior made it resemble some other type of stone not found locally and perhaps therefore significant. Along these lines, the presence of artifacts in the New Guinea Highlands that resemble stemmed tools, but made from di¤erent raw materials (e.g., Bulmer 2005:432 type 3c), is provocative, although in the absence of good chronology for the New Guinea examples, one might well ask who was imitating whom. Rath and Torrence (2003) have suggested that the numerous stages of production necessary for the manufacture of stemmed tools might have contributed to their social significance in several ways. Firstly, the additional care and e¤ort invested in their manufacture may have added to their ‘‘value.’’ Secondly, their production may have required special skills that were not widely shared in the popu- lation, therefore requiring the creation of social relationships with one or more knowledgeable knappers. Thirdly, the staged reduction sequence provided opportunities for moving the tool between several producers or craft specialists located in di¤erent places. Possession of the final tool would therefore provide concrete proof that a series of social relations had been successfully negotiated. Hammer- dressing might then represent an extra step in production that required additional skills or persons and therefore further enhanced the value of these objects. A final suggestion is that this stage of stemmed tool production o¤ered opportunities for ‘‘craft-as-performance’’ as described by Carter (2007:94). The making of a tool could have been a ritualized act, especially since hammering a fragile obsidian tool without breaking it was a very risky business. One could imagine that the rhythm of the hammering and the associated sounds were conducive to a public performance of skills incorporated into a ceremony. 138 asian perspectives . 48(1) . spring 2009

Replication Experiment

To evaluate the Rath and Torrence hypotheses, Nina Kononenko conducted a replication experiment aimed at assessing the time and skill requirements and functional advantages of hammer-dressing. In the first stage of manufacture, a roughly circular, thick flake with a large bulb of percussion was struck from a core of New Britain obsidian using direct percussion. Although not strictly a Kom- bewa flake, the blank resembled many stemmed tools in terms of its bilateral symmetry and longitudinal cross section. The ventral and dorsal surfaces of the flake intersected to form a relatively thick and sturdy distal edge. Next, a combination of invasive and steep bifacial percussion was applied to the bulbar end of the flake to create the two notches that delineate the stem and create its roughly triangular cross section. Four heavy river cobbles of relatively soft material were used as hammerstones to make the stemmed tool. The presence of a prominent ridge creating the thick, triangular cross section of the stem and the sharp, fresh flake scars made the stem di‰cult to grasp. Addi- tionally, any movement of the tool within a haft would have torn it and eventu- ally destroyed a wooden collar or wrapping. One way to avoid this problem would be to smooth the surface of the stem and at the same time convert the cross section into a round or oval shape more easily hafted. To achieve these goals, hard hammer blows were applied directly to the surface of the stem and the notches. During the hammer-dressing, as much as possible of the tool rested on a soft wood anvil to prevent unequal stress that could cause breakage at the junction of the stem and the body of the flake. During much of the experiment the sharp distal edge of the flake was covered by a soft animal skin to protect the hand holding the tool. Of eight hammerstones tested, two were preferable. Both of these river cobbles had a narrow edge and one had a relatively pointed shape. It was found that limiting the amount of surface that came in contact with the obsidian surface increased the accuracy of the hammering. Hammer-dressing was restricted to the area of the stem with deep flake scars. The blows were positioned carefully and not applied too forcefully so as to avoid breakage. During this process, the sharp edges of the stem were replaced by small round impact cones created by the hammering (Fig. 12). Although on the small scale these create an irregular surface, it is relatively level and smooth to the touch. During this process microflakes were also removed, although these were not intended. Despite care taken to anchor the tool, a section of the weakest part of the edge, representing the corner just above the notch on one side, su¤ered vibrations and broke o¤ due to end shock (Fig. 12). After 40 minutes, the cross section of the stem had been rounded and the sharp edges removed. At this stage the stem could be held comfortably in the hand. The conical impact fractures created on the experimental tool and the hammer-dressed surfaces of the archaeological artifacts are roughly similar, although those created on the experimental surface are less distinct and more highly fractured and crushed, possibly due to the use of softer hammerstones with a more rounded surface and/or the application of a higher density of blows (Fig. 13). The experiment demonstrated that hammer-dressing is an e¤ective way to smooth contours and reshape a flaked stem, but also that the procedure is torrence et al. . mid-holocene social interaction in melanesia 139

Fig. 12. Final form of hammer-dressed obsidian stemmed tool from replication experiment. Note breakage on the left-hand tip due to end shock during manufacture. Scale bar is 2 cm. (Photo by N. Kononenko) relatively time-consuming and demands appropriate equipment (hammerstones, anvil, hand protection) that must be used with relative precision. An experienced person using a better hammerstone might complete the task more quickly, but pecking still takes more time than flaking. It is also clear from the replication experiment that care is needed to create the desired surface without damaging the tool. The likelihood of breakage explains why all the hammer-dressed stemmed tools are robust. Hammering is a very risky strategy when there are protruding edges or tips and favors stems formed from the thickest (bulbar) end of the flake blank. In summary, the replication experiment provides support for hypotheses that consider the role of the handle, energy and skill inputs, and multiple stages of production, since all are required. Freshly flaked obsidian is sharp, so dulling a potential handle reduces risks of injury and damage to a haft or handle. The procedure also requires extra time, care, skill, and perhaps practice. Although discrimination among the potential reasons for the use of hammer-dressing on stemmed tools will require further research and more data about how the tools were used and exchanged in the past, the use-wear analysis and replication experiment are helpful in suggesting future directions. 140 asian perspectives . 48(1) . spring 2009

Fig. 13. Close-up views of impact fractures on hammer-dressed obsidian stemmed tools. Clockwise from top left: FAP M200; FAP M702; FAP M408; experimental tool. Scale bar is 1 cm. (Photos by R. Torrence) innovation, imitation, and interaction Although it is possible that the combination of the two unusual and highly de- manding techniques, hammer-dressing and Kombewa flakes, was invented inde- pendently at the two widely separated obsidian source areas on Garua and Lou Islands (in New Britain and Manus, respectively; Figs. 1, 4), in our view such a coincidence is highly unlikely. The more plausible scenario is that the similar- ities are the result of significant contact between people living on two distant islands. Perhaps the precise form of the artifact was copied by knappers from one areaafterviewingit,eitherattheotherproductioncenteroratanyofthewidely distributed places where stemmed tools may have been exchanged, as represented by the current find spots (Fig. 1). The Kombewa technique is so complex and dif- ficult that we feel imitation is also highly unlikely. It is most plausible that the replication of identical technologies in two di¤erent places was the outcome of prolonged interaction. We suggest that the use of similar methods to make obsidian stemmed tools was either because (1) people shared ideologies about the role and meaning of the tools, or that (2) competing groups on Garua and Lou used the same forms to gain access to exchange with third parties. torrence et al. . mid-holocene social interaction in melanesia 141

Although limited, there is some archaeological evidence that obsidian stemmed tools may have been made on Lou Island. Two Type 2 stemmed tools and a curved stem from an unknown type have been reported from the nearby mainland of Manus (cf. Araho et al. 2002: Table 1; Bu¨hler 1935:21, 1946–1949:230; Casey 1939:147, Fig. 6; Nevermann 1934:341, Fig. 193; Ohnemus 1996:363, Fig. 433.10), but none has been found near the obsidian sources on Lou Island. Another problem is that Lou Island obsidian is rare in the mid-Holocene archaeological record (e.g., White 1996) and none has been previously found outside Manus province. In addition, only four artifacts from mid-Holocene levels at the Pamwak site on Manus have been characterized to Lou Island sources and one of these is not from the Umleang source (Fredericksen 1997:72). Since artifacts with stems or tangs (but not made on Kombewa flakes) were made on Lou Island beginning in the late Holocene and continuing up to the recent period (Friedericksen 1994, 2000), Type 2 stemmed tools might represent their mid-Holocene precursors. Furthermore, the current lack of evidence for stemmed tools on Lou Island during the mid-Holocene could be explained by the paucity of contemporary contexts (Kennedy 2002). For example, excavations at the site of Umleang, where obsidian outcrops match the chemical composition of the obsidian used in the Biak tool, have not reached the mid-Holocene landscape, which is buried by deep layers of recent volcanic ash that overlie most of the island (e.g., Ambrose 1988, 1998; Ambrose et al. 1981; Fredericksen 2000). Although perhaps unlikely because of the large distances involved, it is not im- possible that people from Lou Island took obsidian to Garua Island where only certain people had the skills, knowledge, and right to make hammer-dressed tools. We already know that Kutau-Bao obsidian from mainland New Britain was transported to Garua Island and converted into Type 1 stemmed tools (Rath and Torrence 2003), and a flake (FAP M228) found on Garua and sourced to Kutau-Bao has been hammer-dressed (Table 3). In addition, since hammer- dressing is absent from the relatively large collections of stemmed tools from mainland New Britain sites at FRL and FCH (e.g., Araho 1996), it is possible that the technique was restricted to Garua Island. In this regard, Nevermann’s (1934) report that in historic times access to the recent obsidian quarries at Umleang and knowledge of blade production was tightly controlled is relevant. The geographical location of the highly skilled knappers in historic Manus, however, was not restricted to the obsidian sources themselves, so this provides a pre- cedent for raw material moving to producers located elsewhere (Fullagar and Torrence 1991:116). Alternately, craft specialists from Garua may have been brought to Lou Island to make stemmed tools by sponsors seeking to enhance their prestige, possibly as part of the performance within a ceremony (cf. Carter 2004, 2007). If so, then these tools may have had a very special value either in social or utilitarian terms. The use-wear results indicate that at least one hammer-dressed tool was used for cutting a large animal, which could include its use in warfare. Perhaps this tool type had special ideological significance within a ceremonial or warfare context. In addition, control over knowledge and skills suggests intense social interaction and possible competition for status, since individuals must have successfully negotiated and manipulated social relationships to obtain raw materials and finished goods (Rath and Torrence 2003). 142 asian perspectives . 48(1) . spring 2009

Further archaeological research, particularly on Lou Island, is necessary to de- termine the precise social relationships implied by the Biak and Garua hammer- dressed obsidian artifacts. At this stage, however, we conclude that independent innovation in two di¤erent places or direct copying are unlikely. Instead, the alternatives—employing craft specialists, or replication following training in this relatively sophisticated technology—all demand signiﬁcant interaction between individuals from Garua and Lou Islands.

interaction spheres Having considered the unusual geological source and technique of hammer- dressing, we now turn to the third important attribute of the Biak artifact: its geographical location. Given the information at hand, the most parsimonious conclu- sion is that the Biak tool was derived from archaeological deposits on Biak that were disturbed by Japanese building activities during the war. Another possibility is that the tool was taken to Biak from Manus by a Japanese soldier as a souvenir, but this scenario is quite unlikely given the history of Japanese troop movements during World War II. On February 29, 1943, the Allies attacked Japanese posi- tions on Manus. Heavy fighting ensued and by May 18, 1943, the fighting was over and 3,317 of the original 4,600 Japanese stationed there had been buried (Dexter1961:795–797).Itisdoubtfulthatanysurvivorsplayedafurtherpartin the war and therefore unlikely that a Japanese soldier fighting on Manus transported the stemmed tool to Biak. If the hammer-dressed stemmed tool was indeed deposited in the Biak cave during prehistory, it had traveled c. 1200 km from the obsidian source on Lou Island. This new find could therefore markedly increase the geographic scale of social relations represented by obsidian stemmed tools to a region extending from WestPapuatosouthernBougainville,adistanceofc.2200km(Fig.1).Theloca- tion of the Biak tool also means that we should seriously reconsider the possibility of social relations between Melanesia and regions to the west during the mid- Holocene. Swadling (1996:15, 51–53) proposed wide-scale interaction beginning in the early Holocene between New Guinea and islands to the west, but until recently there has been very little concrete evidence to support her hypothesis. Drawing on previous work (Araho et al. 2002; Golson 2001, 2005; Swadling and Hide 2005; Torrence 2003, 2004b), Specht (2005:387) proposed that obsidian stemmed tools are ‘‘a highly visible archaeological indicator’’ ‘‘of long-standing social connections’’ during the mid-Holocene. The distance over which the tools traveled from the obsidian sources in New Britain and Manus is so considerable that it is di‰cult to imagine that everyone who possessed one was actively en- gaged within the same ceremonial system of exchange and status, although given our discussion of interaction between the obsidian sources of Garua and Lou, this suggestion requires serious research. More important than the overall distribution of extant stemmed tools, the sharing of the Kombewa technology and hammer-dressing required contact be- tweenthepeoplewhomadetheBiakandGaruaIslandobsidianstemmedtools and these networks must have stretched over areas larger than those recorded by ethnography for recent Melanesian trading systems. The overall spatial pattern of torrence et al. . mid-holocene social interaction in melanesia 143 stemmed tools, however, may not necessarily have resulted from direct communication among all the areas where this tool type has been found. The artifacts may have circulated through a di¤use network of links rather than through a single, organized exchange system. For example, White (1996) has shown that the archaeological distribution of obsidian in the recent period is not restricted to the area encompassed by exchange systems described by ethnographers. Instead, recent obsidian artifacts are spread over a much larger region than self-defined exchange systems because social relationships rarely have tight boundaries, exchange systems often overlap, and objects frequently leak out into a wider social sphere (cf. Specht 2005:385–386). In this regard, the extensive amount of reworking around the blade of the Biak tools is relevant and may signify a long and complex history of use and meanings as it was exchanged and moved through space and time. Stone mortars and pestles and shell artifacts dating to the mid-Holocene have also been used as evidence for wide-scale social interaction within mainland New Guinea and between the islands and the mainland (Golson 2005 :484; Swadling 2005; Swadling and Hide 2005; Swadling et al. 2008). The translocation of various animals to Manus and New Britain is another indication of links between these islands and the New Guinea mainland (White 2004). Perhaps it was on the mainland that people from Manus and New Britain came into contact and then exchangedorborrowedknowledgeandskills relevant to stemmed tools or com- peted for status. Whatever scenario is established by future research, it is clear that the mid-Holocene was a period in which ideas and possibly people were regularly moving over long distances.

implications for pacific prehistory In contrast to recent obsidian exchange systems through which unmodified blocks or small pieces of obsidian were distributed (e.g., Specht 1981), the mid-Holocene stemmed tools are a finely worked product with recognizable shapes. As they passed between hands across social and language boundaries, these distinctive artifacts had the potential to carry various meanings and information with them. The exchange of these tools therefore must have established routes along which a range of ideas, ceremonies, dances, etc., as well as concrete objects could have flowed, possibly as far as from Biak to Bougainville. If people from Manus and New Britain were in regular communication, as implied by our analyses of the hammer-dressed tools, then the spread of goods and ideas might have been relatively rapid. Given the location of Biak, on the edge of Southeast Asia, it is also possible that Melanesia had extensive contacts to the west from at least the mid-Holocene. The existence of a large-scale mid- Holocene interaction sphere or even a loose network of social relations through which goods like stemmed tools, social practices, and ceremonies involved with the stone mortars and pestles were already circulating, created a mechanism for the borrowing and circulation of ideas from the west. If so, then the subsequent adoption and spread of Austronesian languages, Lapita style pottery, pigs, etc., did not necessarily require human migrations. As we have shown, by at least the mid- Holocene period, interaction networks distributing elaborate items of material culture had already set the stage for changes later signaled by Lapita style pottery. 144 asian perspectives . 48(1) . spring 2009

Given the limited amount of archaeological research on the mid-Holocene, the small number of stemmed tools probably represents the tip of the iceberg. As yet the exact timing and motivation for the long-distance interaction represented by the spatial distribution of hammer-dressed obsidian stemmed tools is unknown and requires sustained research. Possibly rising sea levels destroyed homelands as well as created new opportunities that might have spurred on movements of people through and within the broader Southeast Asian/Melanesian region (e.g., Oppenheimer 1998; Terrell 2004:605–606). What is clear, however, is that further research on social interaction during the mid-Holocene period is likely to yield provocative alternative scenarios for Melanesian prehistory.

acknowledgments

The authors are grateful to John Friede for encouraging the research, lending the stemmed tool for study and permitting the characterization analyses. PIXE-PIGME analyses were funded by AINSE grants to Torrence and Peter White. Glenn Sum- merhayes helped interpret the results, which were made possible by the pioneering work of Roger Bird and assistance from Grahame Bailey. The neutron activation was aided by a grant to Rath from the Greenwell Bequest, University of Sydney. Fieldwork on Garua Island was funded by the Australian Research Council, Austra- lia and Paciﬁc Foundation, and Earthwatch Institute and assisted by the PNG Na- tional Research Institute, PNG National Museum and Art Gallery, West New Brit- ain Provincial Cultural Centre, Kimbe Bay Shipping Agencies Ltd., and sta¤ and workers at Garua Plantation. Thanks to Steve Bullard, Australian War Memorial for information about Japanese troop movements; Christina Hellmich, Curator of the Jolika Collection, de Young Museum, for information; Peter White for comments on the manuscript and help with the ﬁgures; and the anonymous reviewers for useful feedback.

references cited

Ambrose, W. 1988 An early bronze artefact from Papua New Guinea. Antiquity 62 :64–74. 1998 Obsidian dating at a recent age obsidian mining site in Papua New Guinea, in Archaeolog- ical Obsidian Studies: Method and Theory: 205–222, ed. M. Schackley. New York: Plenum Press. Ambrose, W., J. Bird, and P. Duerden 1981 The impermanence of obsidian sources in Melanesia, in Archaeological Studies of Pacific Stone Resources:1–19,ed.F.LeachandJ.Davidson.Oxford:BritishArchaeological Reports International Series, 104. Araho, N. 1996 Obsidian Stemmed Tools from West New Britain, Papua New Guinea. Unpublished M.Phil. Thesis, University of Sydney, Australia. Araho, N., R. Torrence, and J. P. White 2002 Valuable and useful: Mid-Holocene stemmed obsidian artifacts from West New Britain, Papua New Guinea. Proceedings of the Prehistoric Society 68:61–81. Bird, R. 1996 Pacific Obsidian Sources. Unpublished ms. on file at the Australian Museum. Bird, R., R. Torrence, G. Summerhayes, and G. Bailey 1997 New Britain obsidian sources. Archaeology in Oceania 32:61–67. Bu¨ hler, A. 1935. Versuch einer Bevo¨lkerungs-und Kulturanalyse aus den Admiralita¨tsinseln. Zeitschrift fu¨r Ethnologie 67 :1–32. torrence et al. . mid-holocene social interaction in melanesia 145

1946– Steingera¨te, Steinskulpturen und Felszeichnungen aus Melanesien und Polynesien. An- 1949 thropos 41–44:225–274, 577–606. Bulmer, S. 2005 Reflections in stone: Axes and the beginnings of agriculture in the Central Highlands of New Guinea, in Papuan Pasts: Cultural, Linguistic and Biological Histories of Papuan-Speaking Peoples: 387–450, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide. Canberra: Pacific Linguistics. Carter, T. 2004 Transformative processes in luminal spaces: Craft as ritual action in the Throne Room area, in Knossos: Palace, City, State: 273–282, ed. G. Cadogan, E. Hatzaki, and A. Vasilakis. London: British School of Athens. 2007 The theatrics of technology: Consuming obsidian in the early Cycladic burial arena. Anthropological Papers of the American Anthropological Association 17:88–107. Casey, D. 1939 Some prehistoric artifacts from the territory of New Guinea. Memoirs of the National Museum of Victoria XI:143–150. Dexter, D. 1961 The New Guinea O¤ensives. Canberra: Australian War Memorial. Fredericksen, C.F.K. 1994 Patterns in Glass: Obsidian and economic specialisation in the Admiralty Islands. Unpub- lished Ph.D. thesis, Australian National University, Canberra. 1997 Changes in Admiralty Islands obsidian source use: The view from Pamwak. Archaeology in Oceania 32:68–73. 2000 Points of discussion: Obsidian blade technology in the Admiralty Islands, 2100 bp to 50 bp. Bulletin of the Indo-Pacific Prehistory Association 20:93–106. Friede, J 2005 New Guinea Art Masterpieces from the Jolika Collection of Marcia and John Friede. 2 vols. San Francisco: Fine Museum of San Francisco. Friedman, J. 1981 Notes on structure and history in Oceania. Folk 23 :275–295. Fullagar, R. 1992 Lithically Lapita. Functional analysis of flaked stone assemblages from West New Britain Province, Papua New Guinea, in Poterie Lapita et Peuplement: 135–143, ed. J.-C. Galipaud,J.Noumea,ORSTOM. 2005 Residue and use-wear, in Archaeology in Practice: A Student Guide to Archaeological Analysis: 207–234, ed. J. Balme and A. Patterson. London: Blackwell. Fullagar, R., G. Summerhayes, B. Ivuyo and J. Specht 1991 Obsidian sources at Mopir, West New Britain Province, Papua New Guinea, Archaeology in Oceania 26:10–14. Fullagar, R., and R. Torrence 1991 Obsidian exploitation at Umleang, Lou Island, in Report of the Lapita Homeland Project: 113–143, ed. J. Allen and C. Gosden. Occasional Papers in Prehistory 20. Canberra: Department of Prehistory, Research School of Pacific Studies, Australian National Uni- versity. Glascock, M. D., G. E. Braswell, and R. H. Cobean 1998 A systematic approach to obsidian source characterization, in Archaeological Obsidian Studies:15–65,ed.S.Shackley.NewYork:Plenum. Golson, J. 1972 Both sides of the Wallace Line: New Guinea, Australia, Island Melanesia and Asian Pre- history, in Early Chinese Art and its Possible Influence in the Pacific Basin: 533–596, ed. N. Barnard. New York: Intercultural Arts Press. 2001 New Guinea, Australia and the Sahul connection, in Histories of Old Ages: Essays in Hon- our of Rhys Jones: 85–210, ed. A. Anderson, I. Lilley, and S. O’Connor. Canberra: Panda- nus Books. 2005 The middle reaches of New Guinea history, in Papuan Pasts: Cultural, Linguistic and Bio- logical Histories of Papuan-Speaking Peoples: 451–492, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide. Canberra: Pacific Linguistics. 146 asian perspectives . 48(1) . spring 2009

Hayden, B. 1983 Social characteristics of early Austronesian colonisers. Indo-Pacific Prehistory Association Bulletin 4:123–134. Hurcombe, L. 1992 Use Wear Analysis and Obsidian Theory, Experiments and Results. She‰eld Archaeological Monographs 4. She‰eld: J. R. Collis Publications. Kennedy, J. 2002 Manus from the beginning: An archaeological overview, in Admiralty Islands: Art from the South Seas: 17–28, ed. C. Kaufmann, C. Kocher Schmid, and S. Ohnemus. Zurich: Museum Rietberg. Kirch, P. 1997 The Lapita Peoples: Ancestors of the Oceanic World. London: Blackwell. McCarthy, F., E. Bramell, and H. Noone 1946 The Stone Implements of Australia. Australian Museum Memoir IX. Sydney: Australian Museum. Moseley, H. N. 1877 On the inhabitants of the Admiralty islands. JournaloftheAnthropological Institute of Great Britain and Ireland 6:379–429. Nevermann, H. 1934 Admiralita¨ts-Insel. Ergebnisse der Su¨dsee-Expedition 1908–1910, Ethnographie: A. Melanesien. Hamburg: Friederichsen, De Gruyter and Co. Ohnemus, S. 1996 An Ethnology of the Admiralty Islanders: The Alfred Bu¨hler Collection, Museum der Kulturen, Basel. Bathurst: Crawford House Publishing. Oppenheimer, S. 1998 Eden in the East: The Drowned Continent of Southeast Asia. London: Weidenfeld and Nicolson. Petrie, C., and R. Torrence 2008 Assessing the e¤ects of volcanic disasters on human settlement in the Willaumez Penin- sula, Papua New Guinea: A Bayesian approach. Holocene 18:729–744. Rath, P., and R. Torrence 2003 Producing value: Stemmed tools from Garua Island, Papua New Guinea. Australian Archaeology 57 :119–127. Riegelman, H. 1955 Caves of Biak: An American O‰cer’s Experiences in the Southwest Pacific. New York: Dial Press. Sheets, P. 1978 From craftsmen to cog: Quantitative views of Mesoamerican lithic technology, in Papers on the Economy and Architecture of the Ancient Maya: 40–71, ed. R. Sidrys. UCLA Institute of Archaeology Monograph VIII. Los Angeles: University of California at Los Angeles. Smith, R. R. 1953 The Approach to the Philippines. United States Army in World War II: The War in the Pacific, vol. 3. Washington, D.C.: O‰ce of the Chief of Military History. Specht, J. 1973 Prehistory poses many problems. Australian Natural History 17 :445–451. 1974 Lapita pottery at Talasea, West New Britain, Papua New Guinea. Antiquity 48:302–306. 1981 Obsidian sources at Talasea, West New Britain, Papua New Guinea. Journal of the Poly- nesian Society 90 :337–356. 2005 Obsidian stemmed tools in New Britain: Aspects of their role and value in mid-Holocene Papua New Guinea, in Many Exchanges: Archaeology, History, Community and the Work of Isabel McBryde, ed. I. Macfarlane, R. Paton, and M. Mountain. Aboriginal History Mono- graph 11:357–372. Canberra: Australian National University. Specht, J., R. Fullagar, R. Torrence, and N. Baker 1988 Prehistoric obsidian exchange in Melanesia: A perspective from the Talasea sources. Aus- tralian Archaeology 27:3–16. torrence et al. . mid-holocene social interaction in melanesia 147

Spriggs, M. 1997 The Island Melanesians.London:Blackwell. Summerhayes, G. R., J. R. Bird, R. Fullagar, C. Gosden, J. Specht, and R. Torrence 1998 Application of PIXE-PIGME to archaeological analysis of changing patterns of obsidian use in West New Britain, Papua New Guinea, in Archaeological Obsidian Studies:129– 158, ed. S. Shackley. New York: Plenum. Swadling, P. 1996 Plumes from Paradise: Trade Cycles in Outer Southeast Asia and their Impact on New Guinea and Nearby Islands until 1920. Coorparoo, Australia: National Museum of Papua New Guinea in association with Robert Brown and Associates. 2005 The Huon Gulf and its hinterlands: A long-term view of coastal-Highlands interactions, in A Polymath Anthropologist: Essays in Honour of Ann Chowning:1–14,ed.C.Gross,H. Lyons, and D. Counts. Research in Anthropology and Linguistics Monograph 6. Auck- land: University of Auckland. Swadling, P., and R. Hide 2005 Changing landscape and social interaction: Looking at agricultural history from a Sepik- Ramu perspective, in Papuan Pasts: Cultural, Linguistic and Biological Histories of Papuan- Speaking Peoples: 289–328, ed. A. Pawley, R. Attenborough, J. Golson, and R. Hide. Canberra: Pacific Linguistics. Swadling, P., P. Wiessner, and A. Tumu 2008 Prehistoric stone artifacts from Enga and the implication of links between the highlands, lowlands and islands for early agriculture in Papua New Guinea. Journal de la Socie´te´ des Oceánistes: 126–7. Terrell, J. 2004 The ‘‘sleeping giant’’ hypothesis and New Guinea’s place in the prehistory of Greater Near Oceania. World Archaeology 36 :601–609. Torrence, R. 1993 Ethnoarchaeology, museum collections and prehistoric exchange: Obsidian-tipped artifacts from the Admiralty Islands. World Archaeology 24:467–481. 2003 ‘‘Like everywhere you’ve never been’’: Archaeological fables from Papua New Guinea, in Theory, Method and Practice in Modern Archaeology: 287–300, ed. R. J. Jeske and D. K. Charles. Westport, Conn.: Praeger. 2004a Pre-Lapita valuables in Island Melanesia, in A Pacific Odyssey: Archaeology and Anthropology in the Western Pacific. Papers in Honour of Jim Specht: 163–172, ed. V. Attenbrow and R. Fullagar. Records of the Australian Museum, Supplement 29. Sydney: Australian Museum. 2004b Now you see it, now you don’t: Changing obsidian source use in the Willaumez Penin- sula, Papua New Guinea, in Explaining Social Change: Studies in Honour of Colin Renfrew: 115–125, ed. J. Cherry, C. Scarre, and S. Shennan. Cambridge: McDonald Institute for Archaeological Research. 2005 Valued stone: How so?, in Many Exchanges: Archaeology, History, Community and the Work of Isabel McBryde: 357–372, ed. I. Macfarlane, R. Paton, and M. Mountain. Aboriginal History Monograph 11. Canberra: Australian National University. Torrence, R., J. Specht, and R. Fullagar 1990 Pompeiis in the Pacific. Australian Natural History 23:456–463. Torrence, R., and C. Stevenson 2000 Beyond the beach: Changing Lapita landscapes on Garua Island, PNG, in Australian Archaeologist: Collected Papers in Honour of Jim Allen: 324–345, ed. A. Anderson, and T. Murray. Canberra: Coombs Press. Torrence, R., and P. Swadling 2008 Social networks and the spread of Lapita. Antiquity 82:600–617. White, J. P. 1996 Rocks in the head: thinking about the distribution of obsidian in Near Oceania, in Oce- anic Culture History: Essays in Honour of Roger Green: 211-24, eds J. M. Davidson, G. Irwin, B. F. Leach, A. Pawley, and D. Brown. Dunedin: New Zealand Journal of Archaeology Special Publication. 148 asian perspectives . 48(1) . spring 2009

2004 Where the wild things are: Animal translocation in the circum New Guinea Archipelago, in Voyages of Discovery. The Archaeology of Islands: 147–164, ed. S. M. Fitzpatrick. West- port, Conn.: Praeger.

abstract

The widespread distributioninPapuaNewGuineaofobsidianstemmedtoolsdated to the mid-Holocene has led scholars to postulate the existence of large interaction spheres. A newly reported artifact from Biak Island, West Papua provides the stimu- lus for reconsidering the role of this tool type in regional social interaction. The tool was hammer-dressed, a technique unknown for obsidian flaked tools elsewhere in the world and only rarely applied to obsidian artifacts in Melanesia. This new find closely resembles hammer-dressed obsidian stemmed tools from Garua Island, Papua New Guinea, but these are characterized by LA/ICPMS, PIXE-PGME, and INAA to the local Baki and Kutau-Bao obsidian sources in New Britain, Papua New Guinea, whereas the Biak tool is sourced to outcrops on Lou Island in Manus Prov- ince, Papua New Guinea. Hypotheses for functional, symbolic, and social roles of hammer-dressing are explored and evaluated on the basis of replication experiments and use-wear analyses. We argue that the complex and exceptionally rare technologies used for manufacturing hammer-dressed stemmed tools and applied to obsidian acquired from two widely separated obsidian sources substantially add to previous evidence for wide-scale social interaction during the mid-Holocene. The existence of these social networks might also have provided a mechanism for the rapid, extensive spread of innovations like Austronesian languages or Lapita pottery. Keywords: Melanesia, Pacific archaeology, stone tools, obsidian, hammer-dressing, characterization, PIXE-PIGME, LA/ICPMS, instrumental neutron activation analysis.